<p>A plane point cloud segmentation method based on the combination of multiscale hypervoxel segmentation and voxel region growth algorithm is proposed in this article. Firstly, the original point cloud is generated into an initial hypervoxel, which does not need to initialize seed points and does not depend on internal parameters, and the point cloud is segmented into multi-scale hypervoxels. Then, the nearest neighbor method is used to evaluate the average distance of points within high voxels. The neighborhood size is manually adjusted according to the median distance threshold to distinguish between planar and non-planar high voxels. Until all hypervoxels are clearly categorized as scales below a predefined minimal size, non-planar hypervoxels are further split. Subsequently, a region growth approach is employed for fine plane segmentation, whereby points with lower curvature are preferentially selected as seed points, and adjacent units with similar geometric and spectral characteristics are gradually fused to form a unified plane region. The outcomes of the experiments show that the proposed approach works demonstrates superior segmentation performance on both ground laser scanning and UAV point cloud data, which is better than RANSAC, DMNE-RG, BRDS and GEO. Moreover, the Precision, Recall, and F1-score of the proposed approach exceed 0.90. Additionally, it shows enhanced segmentation capabilities for UAV building point clouds characterized by uneven point cloud density and low point cloud quality.</p>